data analysis of coded chats tracking the collaboration activity of a student in different powwows...
TRANSCRIPT
Data Analysis of Coded Chats Tracking the collaboration activity of a student in different powwows
Progress Report, VMT Meeting, Feb. 2nd, 2005
Fatos XhafaVMT Project
February 2nd, 2005. VMT Meeting
Outline
The data considered for this analysis
The research questions addressed
The statistical results Discussion Next steps
February 2nd, 2005. VMT Meeting
The data considered for this analysis Student: AME
Participation in 11 powwows (pow9 ― pow19) Only seven of these powwows (pow9, pow10,
pow12, pow13, pow14, pow15, pow18 ) are coded
Powwows considered: pow9, pow10, pow13, pow14, pow18 The descriptive statistics is required as a previous
step Only three of these (pow9, pow10, pow18) are
part of the initial sample of six powwows
February 2nd, 2005. VMT Meeting
The research questions addressed (I) Q1: Is there any significant difference between the AME's “production” and his groups’ “production”?
Q2: In case of significant difference (from Q1), is there any pattern in the variation (difference) between the AME's and his groups’ production?
Q3: Is there any clustering of the AME's “behavior” – as regards its production– in different powwows?
Q4: What are the correlations between the variables for the AME's “production”? Are those different from what we already know from the
correlations for the sample of six powwows?
February 2nd, 2005. VMT Meeting
The research questions addressed (II) Q5: What are the correlations between the
variables for the AME's “production” and the variables of his groups’ “production”?
Q6: How is the AME’s participation chronologically? Is there any evidence of “experience” due to
repeated participation? Or, any other kind of significant influence of
the repeated participation?
February 2nd, 2005. VMT Meeting
The study of AME's “production” vs. his groups’ “production”
(Social reference, Pbm Solving and Math Move vars are used)
February 2nd, 2005. VMT Meeting
The AME's “production” vs. his groups’ “production” There is a significant difference between the
Percentage of Group's Pbm Solving postings and Percentage of AME's Pbm Solving postings
There is a significant difference between the Percentage of Group's Math Move postings and the Percentage of AME's Math Move postings
There is no significant difference (at 95% confidence level) between the Percentage of Group's Social reference postings and Percentage of AME's Social reference postings but a significant difference at 90% confidence level
February 2nd, 2005. VMT Meeting
Analytically…
Mean N Std. Deviation Std. Error Mean
Pair 1
Percentage of Group's Social reference postings
24.3800 5 7.32236 3.27466
Percentage of AME's Social reference postings
35.9200 5 8.64390 3.86567
Pair 2
Percentage Group's Pbm Solving postings
26.3400 5 6.76705 3.02632
Percentage of AME's Pbm Solving postings
34.0400 5 6.62442 2.96253
Pair 3 Percentage Group's Math Move postings
12.6400 5 6.26123 2.80011
Percentage of AME's Math Move postings
31.1200 5 9.29500 4.15685
Paired Samples Statistics
February 2nd, 2005. VMT Meeting
Analytically…
Paired Differences t dfSig.
(2-tailed)
MeanStd.
DeviationStd. Error Mean
95% ConfidenceInterval of theDifference
Lower Upper
Percentage of Group's Social reference postings - Percentage of AME's Social reference postings
-11.54000 11.52749 5.15525 -25.85327 2.77327 -2.238 4 .089
Percentage Group's Pbm Solving postings - Percentage of AME's Pbm Solving postings
-7.70000 4.37893 1.95832 -13.13716 -2.26284 -3.932 4 .017
Percentage Group's Math Move postings - Percentage of AME's Math Move postings
-18.48000 4.20678 1.88133 -23.70341 -13.25659 -9.823 4 .001
February 2nd, 2005. VMT Meeting
Clustering of the AME's “production” in different powwows
February 2nd, 2005. VMT Meeting
Clustering of the AME's “production” in different powwows The study shows the following clustering
(based on the AME’s “production” in the four dimensions, now including Conversation):
pow9, pow13, pow18 pow10, pow14
February 2nd, 2005. VMT Meeting
Analytically…Cluster Membership
Case NumberPowwow
group Cluster
1 pow9 1
2 pow10 2
3 pow18 1
4 pow13 1
5 pow14 2
Final Cluster Centers
Cluster
1 2
Percentage of AME's Conversation postings
29.57 37.35
Percentage of AME's Social reference postings
30.00 44.80
Percentage of AME's Pbm Solving postings
30.40 39.50
Percentage of AME's Math Move postings
26.27 38.40
Distances between Final Cluster Centers
Cluster 1 2
1 22.575
2 22.575
Number of Cases in each Cluster
Cluster 1 3.000
2 2.000
Valid 5.000
Missing .000
February 2nd, 2005. VMT Meeting
Study of correlations between the variables of the AME's “production”
February 2nd, 2005. VMT Meeting
The correlations between the variables for the AME's “production”Correlations
Percentage of AME's
Conversation postings
Percentage of AME's Social
reference postings
Percentage of AME's Pbm
Solving postings
Percentage of AME's Math Move
postings
Percentage of AME's Conversation postings
Pearson Correlation1 .918(*) .827 .782
Sig. (2-tailed) . .028 .084 .118
N 5 5 5 5
Percentage of AME's Social reference postings
Pearson Correlation .918(*) 1 .821 .858
Sig. (2-tailed) .028 . .088 .063
N 5 5 5 5
Percentage of AME's Pbm Solving postings
Pearson Correlation.827 .821 1 .919(*)
Sig. (2-tailed) .084 .088 . .027
N 5 5 5 5
Percentage of AME's Math Move postings
Pearson Correlation.782 .858 .919(*) 1
Sig. (2-tailed) .118 .063 .027 .
N 5 5 5 5* Correlation is significant at the 0.05 level (2-tailed).
No significant at 0.05 level
February 2nd, 2005. VMT Meeting
The correlations between the variables for the AME's “production” – again, Pow18 is an outlier for Social reference dimension!
Correlations (after removing Pow18)
Percentage of AME's Social
reference postingsPercentage of AME's Pbm
Solving postingsPercentage of AME's Math Move postings
Percentage of AME's Social reference postings
Pearson Correlation
1 .987(*) .954(*)
Sig. (2-tailed) . .013 .046
N 4 4 4
Percentage of AME's Pbm Solving postings
Pearson Correlation
.987(*) 1 .926
Sig. (2-tailed) .013 . .074
N 4 4 4
Percentage of AME's Math Move postings
Pearson Correlation
.954(*) .926 1
Sig. (2-tailed) .046 .074 .
N4 4 4
* Correlation is significant at the 0.05 level (2-tailed).
February 2nd, 2005. VMT Meeting
Regressions (AME’s Soc. Ref vs. AME’s Conversation) graphically:
Linear Regression
25.00 30.00 35.00 40.00
Percentage of AME's Conversation postings
25.00
30.00
35.00
40.00
45.00
Per
cen
tag
e o
f AM
E's
So
cial
ref
eren
ce p
ost
ing
s
pow9
pow10
pow13
pow14
pow18
Percentage of AME's Social reference postings = -12.85 + 1.49 * PConAMER-Square = 0.84
February 2nd, 2005. VMT Meeting
Regressions (AME’s Soc. Ref vs. AME’s Conversation) analytically:Model Summary
Model RR
SquareAdjusted R
Square
Std. Error of the
Estimate Change Statistics
R Square Change
F Chang
e df1df2
Sig. F Change
1 .918(a) .843 .790 2.43614 .843 16.055 1 3 .028
a Predictors: (Constant), Percentage of AME's Social reference postingsb Dependent Variable: Percentage of AME's Conversation postings
Model Sum of
Squares df Mean Square F Sig.
1 Regression 95.284 1 95.284 16.055 .028(a)
Residual 17.804 3 5.935
Total 113.088 4
a Predictors: (Constant), Percentage of AME's Social reference postingsb Dependent Variable: Percentage of AME's Conversation postings
ANOVA(b)
February 2nd, 2005. VMT Meeting
Regressions (AME’s Pbm Solving vs. AME’s Math Move) graphically:
Linear Regression
28.00 32.00 36.00 40.00
Percentage of AME's Pbm Solving postings
20.00
30.00
40.00
Per
can
tag
e o
f AM
E's
Mat
h M
ove
po
stin
gs
pow9
pow10
pow13
pow14
pow18
Percantage of AME's Math Move postings = -12.79 + 1.29 * PPbmSAMER-Square = 0.85
February 2nd, 2005. VMT Meeting
Regressions (AME’s Pbm Solving vs. AME’s Math Move) analytically:
Model Summary
Model RR
SquareAdjusted R Square
Std. Error of the
Estimate Change Statistics
R Square Change
F Change df1 df2
Sig. F Change
1.919(a) .845 .794 4.22240 .845 16.384 1 3 .027
a Predictors: (Constant), Percentage of AME's Pbm Solving postingsb Dependent Variable: Percentage of AME's Math Move postings
Model Sum of
Squares df Mean Square F Sig.
1 Regression 292.102 1 292.102 16.384 .027(a)
Residual 53.486 3 17.829
Total 345.588 4
a Predictors: (Constant), Percentage of AME's Pbm Solving postingsb Dependent Variable: Percentage of AME's Math Move postings
ANOVA
February 2nd, 2005. VMT Meeting
The regression between the AME's variables Soc. Ref, Pbm Solving and Math Move (after removing Pow18)
30.00 35.00 40.00 45.00
Percentage of AME's Social reference postings
20.00
30.00
40.00
Per
can
tag
e o
f AM
E's
Mat
h M
ove
po
stin
gs
pow9
pow10
pow13
pow14
Percantage of AME's Math Move postings = -9.18 + 1.07 * PSRefAMER-Square = 0.91
30.00 35.00 40.00 45.00
Percentage of AME's Social reference postings
28.00
32.00
36.00
40.00
Per
cen
tag
e o
f AM
E's
Pb
m S
olv
ing
po
stin
gs
pow9
pow10
pow13
pow14pow18
Percentage of AME's Pbm Solving postings = 11.44 + 0.63 * PSRefAMER-Square = 0.67
February 2nd, 2005. VMT Meeting
Study of correlations between the AME's vars and his groups’ vars
February 2nd, 2005. VMT Meeting
Correlation between AME’s and his Groups’ vars (Social Reference)
LLR Smoother
15.00 20.00 25.00 30.00
Percentage of Group's Social reference postings
30.00
40.00
50.00
60.00
70.00
80.00
Per
cen
tag
e o
f AM
E's
So
cial
ref
eren
ce p
ost
ing
s
pow9
pow10
pow13
pow14
pow18
No correlation between AME’s Soc. Reference and Groups’
Soc. Reference!
February 2nd, 2005. VMT Meeting
Correlation between AME’s and his Groups’ Math Move (significant correlation)Correlations
Percentage Group's Math Move postings
Percentage of AME's Math Move
postings
Percentage Group's Math Move postings
Pearson Correlation 1
.927(*)
Sig. (2-tailed).
.023
N 5 5
Percantage of AME's Math Move postings
Pearson Correlation .927(*)
1
Sig. (2-tailed).023
.
N 5 5
* Correlation is significant at the 0.05 level (2-tailed).
Linear Regression
20.00 30.00 40.00
Percantage of AME's Math Move postings
5.00
10.00
15.00
20.00
Per
cen
tag
e G
rou
p's
Mat
h M
ov
e p
ost
ing
s
pow9
pow10
pow13
pow14
pow18
Percentage Group's Math Move postings = -6.79 + 0.62 * PMathAMER-Square = 0.86
February 2nd, 2005. VMT Meeting
Correlations between AME’s and his Groups’ Pbm Solving (almost significant, 0.057 level)
20.00 24.00 28.00 32.00
Percentage Group's Pbm Solving postings
28.00
32.00
36.00
40.00
Per
cen
tag
e o
f AM
E's
Pb
m S
olv
ing
po
stin
gs
pow9
pow10
pow13
pow14pow18
Percentage of AME's Pbm Solving postings = 13.76 + 0.77 * PPbmSoGrR-Square = 0.62
Percentage of AME's Pbm
Solving postings
Percentage Group's Pbm
Solving postings
Pearson Correlation
Percentage of AME's Pbm Solving postings
1.000 .786
Percentage Group's Pbm Solving postings
.786 1.000
Sig. (1-tailed) Percentage of AME's Pbm Solving postings
. .057
Percentage Group's Pbm Solving postings
.057 .
N Percentage of AME's Pbm Solving postings
5 5
Percentage Group's Pbm Solving postings
5 5
February 2nd, 2005. VMT Meeting
Study of AME’s participation chronologically[“Who shapes the collaboration?”]
February 2nd, 2005. VMT Meeting
AME’s participation chronologically (Pbm Solv.)
04/15/200404/25/2004
05/06/200405/09/2004
05/27/2004
Date of the powwow
15.00
20.00
25.00
30.00
35.00
40.00
45.00 Percentage of AME's Pbm Solving postings
Percentage Group's Pbm Solving postings
High “agreement” in the shape of the two
curves!!
[Not surprising, as we know the correlations between the AME’s vs. his groups’ vars]
February 2nd, 2005. VMT Meeting
AME’s participation chronologically (Math Move)
04/15/200404/25/2004
05/06/200405/09/2004
05/27/2004
Date of the powwow
10.00
20.00
30.00
40.00
Percantage of AME's Math Move postings
Percentage Group's Math Move postings
High “agreement” in the shape of the two
curves!!
[Not surprising, as we know the correlations between the AME’s vs. his groups’ vars]
February 2nd, 2005. VMT Meeting
AME’s participation chronologically (Soc. Ref)
04/15/200404/25/2004
05/06/200405/09/2004
05/27/2004
Date of the powwow
10.00
20.00
30.00
40.00
50.00 Percentage of AME's Social reference postings
Percentage of Group's Social reference postings
No “agreement” in the shape of the two
curves!!
[Not surprising, as we know the correlations between the AME’s vs. his groups’ vars]
February 2nd, 2005. VMT Meeting
Discussion…back to the research questions Q1: Is there any significant difference between
the AME's “production” and his groups’ “production”? Yes, as regards the Pbm Solving and Math Move. No, as regards the Social reference, though there is a
considerable difference (significant at 90% conf. level). Q2: In case of significant difference (from Q1), is
there any pattern in the variation between the AME's and the groups? Yes, the variation follows a pattern-magnitude difference
(agreement curves). The AME’s “production” is always above the groups’
“production”, in the three dimensions! This might be indicating that AME is a “high ability” student.
February 2nd, 2005. VMT Meeting
An excerpt from AME’s “production” (Pow10, Conversation and Math Move dimensions)
AMEFIRMCPMFP
Handle
Pies show percents
35.48%
22.58%
41.94%
Agree Critique Disagree Explain
Elaborate Extend Follow No Code
Offer Regulate Respond Request
Repair typing State Setup
61.54%30.77%
7.69%
66.67%
33.33%46.67%
6.67%
40.00%
6.67%
38.46%
61.54%
29.17%
6.25%56.25%
8.33%
26.19%
61.90%
11.90%
100.00%
47.22%
1.39%
51.39%63.64%
9.09%
27.27% 33.63%
19.47%24.78%
22.12%39.02%
13.41%
34.15%
13.41%
44.44%
33.33%
22.22%
45.07%
12.68%
25.35%
16.90%
66.67%
20.00%
13.33%
Pie chart of Conversation dimension sliced by Handle
AMEFIRMCPMFP
Handle
Pies show percents100.00%
Algebraic expr. Geometric expr. Import new math
Import & apply new math Numeric computation
36.14%
1.20%
57.83%
4.82%
80.00%
20.00%
57.14%
42.86%50.00%
6.67%
43.33%
Pie chart of Math Move dimension sliced by Handle
February 2nd, 2005. VMT Meeting
Discussion…back to the research questions Q3: Is there any clustering of the AME's
“behavior” – as regards its production– in different powwows? Yes, the data shows a clustering into two groups
of powwows (pow9, pow13, pow18 and pow10, pow14)
We know that pow18 and pow10 fall into different clusters, is the same reason for AME’s clustering?
What is what makes the AME’s behavior to be clustering in this way?
February 2nd, 2005. VMT Meeting
Discussion…back to the research questions Q4: What are the correlations between the
variables for the AME's “production”? Are those different from what we already know from the correlations (the sample of six powwows)? We have now a new correlation, between Conversation
and Social reference, namely a strongly positive correlation, .918)
The correlation between the Pbm Solving and Math Move is as in the case of the sample (strongly positive correlation, .919)
The correlation between Social Ref. and Pbm Solving and Social Ref. and Math Move is reversed (positively correlated)!!
Why? How does the AME’s Soc. Ref activity unfolds in way that it is positively correlated to the Pbm Solving and Math Move?
February 2nd, 2005. VMT Meeting
Why positive correlation between Social Ref. & Pbm Solving and Social Ref. & Math Move for AME?
SOCIAL REFERENCEPOSTINGS
OFF TOPIC POSTINGS(POSTINGS NOT CODED
IN THE PBM SOLVING DIMENSION)
POSTINGS CODED IN THE PBM SOLVING
DIMENSION
These postings (could) negatively
influence the production of pbm
solving
These postings DO NOT influence negatively the
production of pbm solving
February 2nd, 2005. VMT Meeting
This leads to…
Hypothesis:
More “off topic” postings in the social
reference implies more negative influence
of social reference in the production
of pbm solving
February 2nd, 2005. VMT Meeting
Why positive correlation between Social Ref. & Pbm Solving and Social Ref. & Math Move for AME? (POWWOW10)
SOCIAL REFERENCEPOSTINGS(166 overall)
OFF TOPIC POSTINGS(POSTINGS NOT CODED
IN THE PBM SOLVING DIMENSION)
POSTINGS CODED IN THE PBM SOLVING
DIMENSION
AME: 13/81=16%
AME: 68/81=84%
FIR: 14/34=42%
MCP: 8/51=15%
FIR: 20/34=58%
MCP: 43/81=85%
February 2nd, 2005. VMT Meeting
Why positive correlation between Social Ref. & Pbm Solving and Social Ref. & Math Move
for AME?
AMEFIRMCP
Handle
Pies show percents
60.47%
2.33%
37.21%
Collaboration group Collaboration individual Greet Home
Interest Identify other Identify self Risk-taking
School Sustain climate
44.12%
20.59%
35.29%50.00%
25.00%
25.00%
100.00%
83.33%
16.67%
50.00%50.00%
75.00%
25.00% 25.00%
75.00%
30.77%
30.77%
38.46%
58.33%25.00%
16.67%
Pie chart of Social reference sliced by Handle
Most of the “Identify self” are from AME”
(showing his own abilities/math skills)
Most of the “Group refs” are by AME”(addressing the
group)
Most of the “Group refs” are by AME”(addressing the
group)
February 2nd, 2005. VMT Meeting
Discussion…back to the research questions Q5: What are the correlations between the
variables for the AME's “production” and the variables of his groups’ “production”? There is No correlation between AME’s Soc.
Reference and Groups’ Soc. Reference! There is a significant correlation between AME’s
and his Groups’ Math Move There is (almost) a significant correlation between
AME’s and his Groups’ Pbm Solving
February 2nd, 2005. VMT Meeting
Discussion…back to the research questions Q6: How is the AME’s participation chronologically? Is
there any evidence of “experience” due to repeated participation? Or any kind of significant influence of the repeated participation? From the analysis we see that repeated participation doesn’t
mean that the participant’s production is greater and greater… It could be indicating that “no matter how experienced or high-
ability you are, your production is by large a matter of the group”!!
How does the group’s “circumstances/conditions” influence the participants’ production (even in the sense of “producing” more or less?)
Yet, an experienced participant can show/keep a pattern of “producing” more than the group does! This is the case of AME
February 2nd, 2005. VMT Meeting
Next steps From statistical perspective
Complete the same computations by including the 2 powwows (and later with the rest of 4 powwows that are not coded yet)
Refine the analysis (second level analysis including subcategories of the dimensions)
Consider another student (FIRSUNSHINE?) that could be another interesting case (“low-ability”?) preferably from the AME’s powwows
From thread-based analysis Analyze the AME’s participation by looking at the “patterns” produced by
AME Analyze the AME’s participation by looking at the threads in which he is
involved From interaction based / CA perspective
Analyze how does the AME’s activity unfold sequentially in different powwows, especially his social reference production
How does the group’s “circumstances/conditions” influence the participants’ production (even in the sense of “producing” more or less?)
How does the type of organization (exposition vs. exploration) influence the AME’s production?
How can, from a CA perspective, be explained the AME’s clustering?